the one about unsigned integers with one bit missing would be trivial to implement as a library in C++ with no significant downside. all you have to do is make a class wrapping a signed int and put debug checks for if the high 1 bit is set behind an if constexpr in operator= and the copy constructor. in most other languages this would bring a big performance penalty, but this is one thing that C++ is actually very good at.
This is incorrect. For one thing, you're going to need to overload all the operators that mutate the integer in place, e.g. operator++, operator +=, operator -=, operator *=, shifts, etc. And those checks can be quite expensive. For example, your code is probably littered with for loops that are implemented using operator++ or operator+=, and that means on every loop iteration you need to check for overflow, which is expensive if the loop body is simple. GCC and Clang already implement -ftrapv which does something similar (it adds compiler checks at all places where signed integers can overflow and trap if overflow occurs). I've used -ftrapv in debug builds but for most programs you don't want it in non-debug builds because the overhead is too high.
